May 23, 2017

Thank you

Jay Kaufman, Kathryn Morrison, and Sam Harper

The data

  • 40 "states": 39 states + DC
  • 45 years: 1969–2013
  • 19 age groups: <1, 1-4, 5-9, …, 85+
  • 2 genders
  • 2 races (black, white)
  • 6 major causes of deaths: cardiovascular disease, cancer, non-communicable, communicable, injuries, all other causes
  • = 820,800 rows of data

The analysis

  • Estimated smoothed mortality rates using a Bayesian time series model with Poisson likelihood
  • Used a truncated Poisson likelihood when number of deaths were between 1 and 9
  • 1,000 posterior samples. For each:
    1. Create a life table for every strata
    2. Calculate the difference in life expectancy between blacks and whites
    3. Decompose the difference by age and cause of death
    4. Calculate age-standardized mortality rates
  • Calculate the mean, 5th and 95th percentiles for all estimates
  • Compute national trend using random-effects meta regression

The output

The app

The app

The app

Male black-white life expectancy gap

Female black-white life expectancy gap

National patterns in cause contributions to the gap

Three examples

CVD in men

Cancer in women

Injury in women

A tale of two states

Tennessee

New York

Similarities & differences

  • Substantial contribution of infant mortality to the gap in 1969
  • Most deaths from all other causes are during infancy
  • Similar pattern of injury contributions by gender in 1969, but larger contribution in Tennessee
  • Similar pattern of CVD contributions in 1969 and 2013, but Tennessee started worse off
  • Ditto for cancer
  • Injury plays important role in Tennessee in 2013 with opposing effect directions in men and women

Get in touch